Downloads · 30 days
4.8K
21% of all-time downloads
gemma-4-E4B-it-heretic-QAT-GGUF
SC117/gemma-4-E4B-it-heretic-QAT-GGUF
gemma-4-E4B-it-heretic-QAT-GGUF is a machine learning model from SC117. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; display: flex; flex-direction: column; gap: 20px; margin-bottom: 30px;" <div style="border: 1px solid cbd5e1; border-radius:…
All-time downloads
23.2K
Public
Repo size
5.2 GB
Likes
28
Public
Repo makeup
Click a slice to open those files.
.gguf5.2 GB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Jun 8, 2026
- Updated
- Oct 3, 2026
- SHA
- 754bb5b5
Base models
Model card
From the Hugging Face model README
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; display: flex; flex-direction: column; gap: 20px; margin-bottom: 30px;">
<div style="border: 1px solid #cbd5e1; border-radius: 16px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 8px rgba(0,0,0,0.04);">
<div style="background: linear-gradient(135deg, #f59e0b 0%, #d97706 100%); padding: 28px 24px; color: white; text-align: center; position: relative; overflow: hidden;">
<div style="position: absolute; top: -30px; right: -30px; width: 120px; height: 120px; background: rgba(255,255,255,0.1); border-radius: 50%;"></div>
<div style="position: absolute; bottom: -20px; left: 40px; width: 80px; height: 80px; background: rgba(255,255,255,0.08); border-radius: 50%;"></div>
<h1 style="margin: 0; font-size: 24px; font-weight: 700; position: relative; z-index: 1;">⚡ Gemma 4 E4B Heretic QAT — Q4_0 GGUF</h1>
<p style="margin: 8px 0 0 0; font-size: 15px; opacity: 0.95; position: relative; z-index: 1;">Heretic ARA · QAT-Lossless Q4_0 · 4 GB · Mobile-Ready</p>
<p style="margin: 8px 0 0 0; font-size: 14px; position: relative; z-index: 1;">
<a href="https://huggingface.co/SC117/gemma-4-E4B-it-heretic-QAT-GGUF/blob/main/README_zh.md" style="color: #ffffff; text-decoration: none;">📖 中文文档</a></p>
</div>
<div style="padding: 16px 24px; background: #f8fafc; border-bottom: 1px solid #e2e8f0; display: flex; flex-wrap: wrap; gap: 8px;">
<span style="background: #fef3c7; color: #92400e; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">Q4_0</span>
<span style="background: #dbeafe; color: #1e40af; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">4.5B Effective</span>
<span style="background: #fce7f3; color: #9d174d; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">Heretic Uncensored</span>
<span style="background: #d1fae5; color: #065f46; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">4 GB</span>
<span style="background: #ede9fe; color: #5b21b6; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">QAT Weights</span>
<span style="background: #e0f2fe; color: #075985; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600;">📱 Mobile</span>
</div>
<div style="padding: 20px 24px;">
<p style="margin: 0; font-size: 14px; color: #334155; line-height: 1.7;">
Uncensored version of <a href="https://huggingface.co/google/gemma-4-E4B-it" style="color: #c2410c; text-decoration: none;">Google Gemma 4 E4B IT (QAT)</a>, processed with <a href="https://github.com/p-e-w/heretic" style="color: #c2410c; text-decoration: none;">Heretic</a> ARA abliteration. Quantized to <b>Q4_0</b> — only 4 GB, perfect for <b>mobile and edge devices</b>.
</p>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #7c3aed 0%, #a855f7 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>✂️</span> Heretic ARA Abliteration Parameters</div>
<div style="padding: 16px;">
<p style="margin: 0 0 8px 0; font-size: 13px; color: #64748b; font-style: italic; border-left: 3px solid #ddd6fe; padding-left: 12px;">Base: <a href="https://huggingface.co/coder3101/gemma-4-E4B-it-qat-q4_0-unquantized-heretic" style="color: #c2410c; text-decoration: none;">coder3101/heretic-QAT</a> · Heretic v1.2.0 · ARA + Row-Norm</p>
<table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px;">
<thead>
<tr style="background: rgba(124,58,237,0.05);">
<th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #5b21b6; font-weight: bold;">Parameter</th>
<th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #5b21b6; font-weight: bold;">Value</th>
</tr>
</thead>
<tbody>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">start_layer_index</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">20</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">end_layer_index</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">36</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">preserve_good_behavior_weight</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">0.5805</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">steer_bad_behavior_weight</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">0.0023</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">overcorrect_relative_weight</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">0.7764</td></tr>
<tr><td style="padding: 8px 10px; color: #334155;">neighbor_count</td><td style="padding: 8px 10px; color: #334155;">13</td></tr>
</tbody>
</table>
<table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px; margin-top: 12px;">
<thead>
<tr style="background: rgba(124,58,237,0.05);">
<th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #5b21b6; font-weight: bold;">Metric</th>
<th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: center; color: #5b21b6; font-weight: bold;">Heretic</th>
<th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: center; color: #5b21b6; font-weight: bold;">Original QAT</th>
</tr>
</thead>
<tbody>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">KL Divergence</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155; text-align: center;">0.0065</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155; text-align: center;">0 (by definition)</td></tr>
<tr><td style="padding: 8px 10px; color: #334155;">Refusals</td><td style="padding: 8px 10px; color: #334155; text-align: center;">5/100</td><td style="padding: 8px 10px; color: #334155; text-align: center;">98/100</td></tr>
</tbody>
</table>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #059669 0%, #10b981 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>🏗️</span> Architecture</div>
<div style="padding: 16px;">
<table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px;">
<tbody>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600; width: 40%;">Base Model</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;"><a href="https://huggingface.co/google/gemma-4-E4B-it" style="color: #c2410c; text-decoration: none;">google/gemma-4-E4B-it</a></td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Parameters</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">4.5B effective (8B with PLE embeddings)</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Architecture</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">Dense with Per-Layer Embeddings (PLE) + multimodal</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Layers</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">42</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Hidden Size</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">2,560</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Context Length</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">128K tokens</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Vocabulary</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">262K, 140+ languages</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #047857; font-weight: 600;">Modalities</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">Text + Image + Audio + Video</td></tr>
<tr><td style="padding: 8px 10px; color: #047857; font-weight: 600;">Quantization</td><td style="padding: 8px 10px; color: #334155;">Q4_0 (matching Unsloth UD-Q4_K_XL layout), b9553 llama-quantize</td></tr>
</tbody>
</table>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #0891b2 0%, #06b6d4 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>📊</span> Quantization Details</div>
<div style="padding: 16px;">
<table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px;">
<tbody>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #0e7490; font-weight: 600; width: 40%;">Format</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">Q4_0 (QAT weights optimized for this precision)</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #0e7490; font-weight: 600;">File Size</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">4 GB</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #0e7490; font-weight: 600;">Effective BPW</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">4.54 (all weights Q4_0, norms F32)</td></tr>
<tr><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #0e7490; font-weight: 600;">Tool</td><td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155;">llama-quantize (b9553, CUDA 13.3)</td></tr>
<tr><td style="padding: 8px 10px; color: #0e7490; font-weight: 600;">QAT Advantage</td><td style="padding: 8px 10px; color: #334155;">Q4_0 + QAT weights achieves 98.5% Top-1 vs 90.9% naive Q4_0 (+7.6%)</td></tr>
</tbody>
</table>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #ea580c 0%, #f97316 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>⚙️</span> Recommended Sampling Parameters</div>
<div style="padding: 16px;">
<table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px; margin-bottom: 16px;">
<tbody>
<tr><td style="padding: 6px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155; font-weight: 600; width: 35%;">General</td><td style="padding: 6px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); color: #334155; font-family: monospace; font-size: 12px;">temp=1.0, top_p=0.95, top_k=64</td></tr>
<tr><td style="padding: 6px 10px; color: #334155; font-weight: 600;">Coding</td><td style="padding: 6px 10px; color: #334155; font-family: monospace; font-size: 12px;">temp=0.6, top_p=0.95, top_k=64</td></tr>
</tbody>
</table>
<p style="margin: 0; font-size: 12px; color: #64748b; line-height: 1.5;">Use <code style="background: #f1f5f9; padding: 2px 6px; border-radius: 4px; font-size: 11px;">--jinja</code> flag with llama.cpp. Disable thinking: <code style="background: #f1f5f9; padding: 2px 6px; border-radius: 4px; font-size: 11px;">--chat-template-kwargs '{"enable_thinking":false}'</code>.</p>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #4b5563 0%, #6b7280 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>📝</span> Usage</div>
<div style="padding: 16px;">
<p style="margin: 0 0 12px 0; font-size: 13px; color: #334155; line-height: 1.7;">Only 4 GB — runs on any modern phone with 6GB+ RAM. Compatible with llama.cpp, LM Studio, Jan, MLC LLM, and mobile GGUF runtimes.</p>
<div style="background: #1e293b; border-radius: 8px; padding: 16px;">
<pre style="margin: 0; font-family: 'SF Mono', 'Fira Code', monospace; font-size: 12px; color: #e2e8f0; line-height: 1.6; white-space: pre-wrap;">llama-server \
-m gemma-4-E4B-it-heretic-QAT-UD-Q4_K_XL.gguf \
--jinja -ngl 99 -c 4096 \
--port 8001</pre>
</div>
</div>
</div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);">
<div style="background: linear-gradient(135deg, #7c3aed 0%, #a855f7 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>🔗</span> Credits</div>
<div style="padding: 16px;">
<p style="margin: 0; font-size: 13px; color: #334155; line-height: 1.7;">
<b>Heretic Abliteration:</b> <a href="https://huggingface.co/coder3101/gemma-4-E4B-it-qat-q4_0-unquantized-heretic" style="color: #c2410c; text-decoration: none;">coder3101</a> · <a href="https://github.com/p-e-w/heretic" style="color: #c2410c; text-decoration: none;">Heretic v1.2.0</a> ARA + Row-Norm<br>
<b>QAT Weights:</b> <a href="https://huggingface.co/google/gemma-4-E4B-it" style="color: #c2410c; text-decoration: none;">Google Gemma 4 E4B IT</a><br>
<b>Quantization Recipe:</b> <a href="https://huggingface.co/unsloth/gemma-4-E4B-it-qat-GGUF" style="color: #c2410c; text-decoration: none;">Unsloth UD-Q4_K_XL</a> (Q4_0 layout)<br>
<b>Quantization Tool:</b> llama.cpp b9553 · <a href="https://github.com/ggerganov/llama.cpp" style="color: #c2410c; text-decoration: none;">GitHub</a><br>
<b>Original Model:</b> <a href="https://huggingface.co/google/gemma-4-E4B-it" style="color: #c2410c; text-decoration: none;">Google Gemma 4 E4B IT</a>
</p>
</div>
</div>
</div>
<!-- SC117_SUPPORT_START -->
<div align="center">
<a href="https://ko-fi.com/sc117">
<img
src="https://huggingface.co/SC117/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF/resolve/main/assets/support-sc117.png"
alt="Support SC117 on Ko-fi — support is optional and all releases remain freely available."
width="100%"
/>
</a>
</div>
<!-- SC117_SUPPORT_END -->
- License
- apache-2.0
- Languages
- en, zh, multilingual
- Created
- Jun 8, 2026
- Updated
- Oct 3, 2026